Prior to 1993, most public assistance programs in the U.S. required applicants to demonstrate not only low incomes but also low assets to qualify for various means-tested programs.  These asset restrictions, not surprisingly, imposed a strong saving disincentive for low-income families (Hubbard, Skinner, & Zeldes, 1995). In 1988, the federal government began permitting states to experiment with new welfare programs.  Some state governments have taken the opportunity to increase the asset limits, in hopes of tempering the saving disincentive found under the old system during the early 1990s.

Few studies have examined how asset limits affects welfare recipients and other low-income people’s saving behavior. Gruber & Yelowitz (1999) and Neumark & Powers (1998) show that Medicaid and SSI are disincentives to saving among those who are likely to eligible for these programs. Using data collected before state welfare reform (1978-1983), Powers (1998) finds that an increase of $1 in asset-limits for AFDC families raises a female head’s saving by 25 cent. To our best knowledge, Hurst & Ziliak (2003) is the only study that examines the effect of post-1992 welfare reforms on saving.  Drawing from the 1994 and 1999 waves of Panel Study of Income Dynamics, they argue that increased asset limits after 1992 have not had a significant impact on saving among families with the highest propensity for receiving welfare (i.e., female heads with children) in both absolute terms and relative to a comparison group of females heads without children.  We are hesitant to believe that Hurst and Ziliak have the final say on this front because they did not take into account the timing of adopting a new asset limit policy and their observation period was not long enough to observe the effect of policy changes. 

In order to fill the gaps in previous study, this paper examines the impact of changes in state-level asset limits on the saving behaviors of low-income women.  Our study combines individual-level behaviors from the Panel Study of Income Dynamics with state-level variation in policy adoption.  We merge this individual-level data with a state-level dataset that captures variation in when asset limit policies were changed and  how they were changed.  The dependent variable of this study is change in liquid asset (savings in bank accounts, CDs, and stocks) between two observation periods.

We use a difference-in-difference analysis to estimate the effect of new asset limits on savings (Neumark and Powers, 1998).  The difference-in-difference estimator identifies the effects of asset limit policy change from the difference—between states that do and do not a new asset limit– in the difference in saving between individuals likely to participate in welfare and individuals unlikely to participate in welfare. 

The results show mixed outcomes. Analyses with continuous measure of saving change indicate that state asset limit policy does not make significant differences in saving among those at high risk of receiving welfare (target population). When a dichotomous measure (positive saving) is used, asset limit policy change has a quite substantial and statistically significant coefficient.